It has long been known that iterative reconstruction algorithms produce different noise characteristics than those produced from the standard filtered backprojection algorithms. This project was undertaken to determine if these noise characteristics would improve the detectability of tumors in positron emission tomography studies using 18-FDG. Initial studies used a novel technique of Ogated replicatesO which permitted actual measurements of noise in patient studies. By comparing the intensity of the signal from tumors to the noise in the tumor region and to the noise in surrounding background regions, it was found that signal to noise improved markedly with use of the expectation maximization algorithm. The results were striking enough to warrant acquiring all further oncology studies using this reconstruction program. The algorithm has been implemented for preliminary clinical use. The preliminary results of the study were accepted for oral presentation, and a full manuscript is under preparation.